What are AI proof SWE skills?

The true "AI proof" skill in my opinion is understanding the problems/pain points/business problem and coming up with a simple solutions to problems

AI is very good at things that can be easily verified with tests/UI changes or building isolated software.

But AI cannot understand and learn the context of the business problem/how different services/pieces of code interacts with each other/why a piece of code exists - the human element of building software

Example: You ask AI to swap ur auth from SSO to email/password. Now what are all the edge cases that can come up?

If your backend expects to store a persons profile pic/ name that SSO gives thats gonna crash

Now you need to handle reset password, forgot password. You need to setup an email client

Example 2: On Taro when u post a question, originally it would be very jarring when u posted and it froze for 2 seconds before the redirect happened. Now it shows a toast banner that says question submitted giving the user the confirmation that it went thru. It's things like that that require understanding ur app that I think will be the edge

All of the things you mentioned are good - I have seen AI struggle with all of those. It can be pretty decent at debugging though as the success criteria there is more clear.

For technical things, I would also add code review. Code review is an extremely contextual process that depends a lot on the engineering cultural and legacy historical decisions of the team (i.e. code quality means different things in different teams). I have tried a couple code review AI bots, and they're all terrible. I can only imagine how terrible it would have been reviewing code on my Instagram team.

Here's another good thread about the topic: "What are product skills and how to develop them in the age of ChatGpt and CopilotX?"

The good news is that you don't need to do anything particularly different in the age of AI. The growth path engineers were already following before ChatGPT took the world by storm, particularly at top tech companies, is still the same. It's just that the pace is accelerated and junior engineers are pressured to advance on that path much faster. Start off with this: https://www.jointaro.com/learning-paths/nail-your-promotion-junior-to-mid-level-l3-to-l4/